CN106469336A - Two-dimension code anti-counterfeit prediction meanss based on BP neural network and method - Google Patents

Two-dimension code anti-counterfeit prediction meanss based on BP neural network and method Download PDF

Info

Publication number
CN106469336A
CN106469336A CN201610843520.8A CN201610843520A CN106469336A CN 106469336 A CN106469336 A CN 106469336A CN 201610843520 A CN201610843520 A CN 201610843520A CN 106469336 A CN106469336 A CN 106469336A
Authority
CN
China
Prior art keywords
barcode scanning
neural network
product
node
output
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201610843520.8A
Other languages
Chinese (zh)
Other versions
CN106469336B (en
Inventor
李志�
陈光锋
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangdong University of Technology
Original Assignee
Guangdong University of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guangdong University of Technology filed Critical Guangdong University of Technology
Priority to CN201610843520.8A priority Critical patent/CN106469336B/en
Publication of CN106469336A publication Critical patent/CN106469336A/en
Application granted granted Critical
Publication of CN106469336B publication Critical patent/CN106469336B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/018—Certifying business or products
    • G06Q30/0185—Product, service or business identity fraud

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • Human Resources & Organizations (AREA)
  • General Business, Economics & Management (AREA)
  • Marketing (AREA)
  • Development Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Data Mining & Analysis (AREA)
  • Finance (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Accounting & Taxation (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • General Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Game Theory and Decision Science (AREA)
  • Health & Medical Sciences (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a kind of two-dimension code anti-counterfeit prediction meanss based on BP neural network and method, this device includes being set up the data access module of false proof database for accessing product data, being used for obtaining the barcode scanning module of barcode scanning data, be used for calling false proof database that the data that barcode scanning obtains is carried out with Treatment Analysis as what signal flow direction was sequentially connected, obtains the data analysis module of attribute character value of product, for building the learning training module of BP neural network algorithm and the algorithm application module that using BP neural network algorithm, barcode scanning product is carried out with false proof prediction；Realization allows elder generation of consumer barcode scanning judge the true and false, then decides whether to buy, and so can solve the problem that consumer must first buy the defect of product ability barcode scanning.

Description

Two-dimension code anti-counterfeit prediction meanss based on BP neural network and method
Technical field
The present invention relates to two-dimension code anti-counterfeit electric powder prediction, more particularly to a kind of Quick Response Code based on BP neural network False proof prediction meanss and method.
Background technology
With the raising of social life level, the requirement more and more higher to product quality for the people, to anti-counterfeiting technology application Demand also accordingly increases.It is wide variety of Antiforge system, product and two dimension both at home and abroad at present based on the Antiforge system of Quick Response Code Code binding, scans Quick Response Code with smart mobile phone, points out authenticity of products situation.Meanwhile, with the popularization of smart mobile phone, mass participation Product false proof will be significantly simpler, and two-dimension code anti-counterfeit systematic difference will be more and more extensive.For current two-dimension code anti-counterfeit system, First, most anti-fraud systems are all to be judged the true and false according to product Quick Response Code by sweeping number of times, such as application documents Dimension code anti-counterfeit method in CN20130617837.6, Quick Response Code first time is swept suggests that certified products, and otherwise prompting personation is produced Product, if identical product is swept by so multiple consumer, judge excessively to think in absolute terms to an authenticity of products；Second, in order to prevent producing Product Quick Response Code is repeatedly swept causes true and false erroneous judgement to be contacted with by lawless person and utilize, and Quick Response Code is placed in interiors of products, so disappears After expense person must buy, ability barcode scanning judges authenticity of products, even fake products, nor predicts in advance.If Quick Response Code bag It is contained in outside product, consumer first barcode scanning can judge the true and false, then determine to buy or do not buy.For this Antiforge system, illegal Molecule can first touch product Quick Response Code, if Quick Response Code is cracked and forges, Antiforge system would become hard to judge the true and false.
BP neural network is one of most widely used neural network model at present.BP network can learn and store substantial amounts of Input-output mode map relation, and the math equation describing this mapping relations need not be disclosed in advance.In application documents It is mentioned in CN201410464900.1 and predict, using BP neural network, the time that task completes, but in the prior art, also Find no the apparatus and method predicting authenticity of products using BP neural network, therefore, based on BP neural network, by barcode scanning The method and apparatus judging authenticity of products is also yet-to-be developed.
Content of the invention
For the deficiencies in the prior art, the purpose of the present invention aims to provide a kind of two-dimension code anti-counterfeit based on BP neural network Prediction meanss and method, in conjunction with false proof database and BP neural network algorithm, predict the probability of authenticity of products, authenticity of products are entered Row rational judgment.
For achieving the above object, the present invention adopts the following technical scheme that：
Based on the two-dimension code anti-counterfeit prediction meanss of BP neural network, including the data access mould being sequentially connected by signal flow direction Block, barcode scanning module, data analysis module, learning training module and algorithm application module；Described BP neural network is by input layer, hidden Constitute containing layer and output layer；
Data access module, sets up false proof database for accessing product data；
Barcode scanning module, for obtaining barcode scanning data；
Data analysis module, for calling false proof database that the data that barcode scanning obtains is carried out with Treatment Analysis, obtains product Attribute character value；
Learning training module, the characteristic value data being obtained by actual product true and false situation as network inputs, make by product False probability, as output, constantly changes the connection weight of network, so that the output of network is constantly close under the stimulation of input Desired output, thus construct BP neural network algorithm；
Algorithm application module, carries out false proof prediction using BP neural network algorithm to barcode scanning product.
As one kind optimization of the present invention, described data access module is used for the converted productss quality information of enterprise, warp Pin business's information, Retailer information, consumer's barcode scanning checking information and packaging 2 D code information are all stored and are set up false proof number According to storehouse.
As one kind optimization of the present invention, described product attribute eigenvalue includes the consistent x1 of barcode scanning information, barcode scanning specifies ground The consistent x2 of point, popularity x3, fraud interests x4, value of risk x5；The information of barcode scanning product information corresponding with false proof database The whether identical composition consistent x1 of barcode scanning information；Barcode scanning place constitutes barcode scanning appointed place one with the variation in designated sale place Cause x2；Described popularity x3 is led to the prescription of product is had differences by living standard difference and constitutes；According to zones of different The punishment degree that product is faked is different, and fraud degree of risk is different, constitutes fraud interests x4；Because different brands production becomes This difference, differently composed value of risk x5 of product profit height.
As one kind optimization of the present invention, the data that barcode scanning obtains is combined by described data analysis module using S type function False proof database is normalized, and it is interval interior, for for BP neural network that the value that barcode scanning is obtained is converted into [0,1] Practise and initial data is provided；Wherein S type function formula is：
Input
Net=x1w1+x2w2+…+xnwn
Output
Wherein xnFor n-th input signal；WnFor weight coefficient.
As one kind optimization of the present invention, described learning training module is according to input signal XiIt is used as by intermediate node Output node, through non-linear transformations, produces output signal Zk, each sample of network training includes input vector X and expectation Output d, the deviation between network output valve Z and desired output d, by adjusting the connection weight of input node and hidden node Value WijBonding strength T and hidden node and output node betweenjkAnd threshold value, so that error is declined along gradient direction, through anti- Multiple learning training, determines the network parameter corresponding with minimum error, i.e. weights and threshold value, training terminates.
Two-dimension code anti-counterfeit Forecasting Methodology based on BP neural network is it is characterised in that include：
Step 1：Set up existing false proof database；
Step 2：Barcode scanning obtains product information；
Step 3：In conjunction with false proof database and barcode scanning information, obtain product attribute eigenvalue；
Step 4：BP neural network learning training；
Step 5：Using BP neural network algorithm, false proof prediction is carried out to barcode scanning product.
As one kind optimization of the present invention, the product attribute eigenvalue described in described step 3 is combined anti-by barcode scanning data Pseudo- data base is normalized to input data, obtains product attribute in [0,1] interval for the numerical value special after calculating Value indicative.
As one kind optimization of the present invention, the specific algorithm of the BP neural network learning training process in described step 4 is such as Under：
(1) assign one group of random initial weight to network, its value is between 0 to 1；
(2) by input data normalized, and determine desired output signal d1 according to actual；
(3) real output value successively according to formula calculating, formula is as follows：
In formula
XiThe output valve of i-th node of-input layer；
YjThe output valve of j-th node of-hidden layer；
WijI-th node of-input layer is to the weight coefficient of j-th node of hidden layer；
WjkJ-th node of-hidden layer is to the weight coefficient of k-th node of output layer；
θjThe internal threshold of j-th node of-hidden layer；
θ1The internal threshold of the 1st node of-output layer；
Z1The real output value of the 1st node of-output layer；
Input layer has N number of input node, and hidden layer has M middle layer node, and output layer has 1 output node.
(4) from the beginning of output layer, add momentum α, wherein 0 ＜ α ＜ 1, reversely adjust weights, its adjustment formula is as follows：
Wjk+ηδkYj→Wjk
Wij+ηδjXi→Wij
δj=Yj(1-Yj)·δ1·Wjk
δ in formula1=(d1-Z1)·Z1·(1-Z1).
(5) total error E, i.e. d are calculated1And Z1Between error amount, if E≤ε, wherein ε terminates for study set in advance Absolute error, study stop, otherwise going in step (3) and recalculate.
As one kind optimization of the present invention, if step-length η is less, pace of learning is slower, and η causes network to occur compared with conference Swing, then add a momentum α, wherein 0 ＜ α ＜ 1 in step (4), that is,
Wjk+ηδkYj+α·ΔWjk→Wjk
Wij+ηδjXi+α·ΔWij→Wij
Δ W in formulajk- double WjkDifference；
ΔWij- double WijDifference.
The beneficial effects of the present invention is：
The present invention updates false proof database data so that prediction effect is more convincing from many aspects；Due to BP nerve The self study of network and adaptive ability are so that predict the outcome relatively accurate；False proof database and BP neural network algorithm combine, The probability of prediction authenticity of products, judges to authenticity of products to avoid thinking in absolute terms, carries out rational judgment, avoid product to a certain extent The erroneous judgement of the true and false；Two-dimension code anti-counterfeit prediction meanss enable to allow elder generation of consumer barcode scanning judge the true and false, then decide whether to buy, and meet Consumer's shopping psychology demand, promotes purchase intention.
Brief description
Fig. 1 is the schematic flow sheet based on the two-dimension code anti-counterfeit prediction meanss of BP neural network for the present invention；
Fig. 2 is the BP neural network illustraton of model of the present invention；
Fig. 3 is the schematic flow sheet based on the two-dimension code anti-counterfeit Forecasting Methodology of BP neural network for the present invention.
Specific embodiment
Below, in conjunction with accompanying drawing and specific embodiment, the present invention is described further：
Embodiment 1
As illustrated in fig. 1 and 2, the invention provides a kind of two-dimension code anti-counterfeit prediction meanss based on BP neural network, including Data access module, barcode scanning module, data analysis module, learning training module and the algorithm application being sequentially connected by signal flow direction Module；Described BP neural network is made up of input layer, hidden layer and output layer.
Data access module, described data access module be used for by the converted productss quality information of enterprise, distributor information, Retailer information, consumer's barcode scanning checking information and packaging 2 D code information are all stored and are set up false proof database, pass through Many-sided moment updates and expands false proof database so that the probability of prediction product fraud is more accurate.Enterprise's converted productss, inspection Test qualified product information and enter line flag, flag information is encrypted with formation product Quick Response Code, supplies after matched packaging code To distributor, simultaneously the information Store such as flag information and product Quick Response Code in false proof database；After distributor receives product, Product is carried out with checking of receiving, checking product packaging code information meets dealer orders, after distributor information and renewal In packaging code Data Enter false proof database；In the same manner, retailer is stored in 2 D code information and retailer when verifying product simultaneously Information, updates and expands false proof database content.
Barcode scanning module, described barcode scanning module also includes barcode scanning device, and barcode scanning device is connected with barcode scanning module by signal, for leading to Cross barcode scanning device barcode scanning and obtain information, for being acquired to the data obtaining after barcode scanning device barcode scanning and storing.Barcode scanning fills Put after Quick Response Code is scanned, the analogue signal that barcode scanning obtains is converted into digital signal by barcode scanning module, is easy to computer Analysis storage barcode scanning information.
Data analysis module, calls at the data that false proof database obtains to barcode scanning in described data analysis module Reason analysis, extracts related data in product database, and barcode scanning information and false proof database are compared in data analysis module Right, final acquisition barcode scanning information is consistent, barcode scanning barcode scanning appointed place is consistent, the product of popularity, fraud interests and value of risk Attribute character value, provides training data for setting up BP neural network；Barcode scanning is obtained by described data analysis module using S type function Data be normalized with reference to false proof database, the value that barcode scanning is obtained be converted into [0,1] interval in, for for BP god Study through network provides initial data；Wherein S type function formula is：
Input
Net=x1w1+x2w2+…+xnwn
Output
Wherein xnFor n-th input signal；WnFor weight coefficient.
Described learning training module is according to input signal XiOutput node is used as by intermediate node, through non-thread deformation Change, produce output signal Zk, each sample of network training includes input vector X and desired throughput d, network output valve Z with Deviation between desired output d, by adjusting connection weight W of input node and hidden nodeijWith hidden node and output Bonding strength T between nodejkAnd threshold value, so that error is declined along gradient direction, through repetition learning training, determine and minimum The corresponding network parameter of error, i.e. weights and threshold value, training terminates；The learning training mistake of BP neural network in the present invention Journey is that the product attribute eigenvalue acquired in the practical situation whether faked by analog equipment is used as BP neural network input, produces Product fraud probability, as output, realizes BP neural network by optimizing the weight coefficient revising input layer, hidden layer and output layer Training；Will the consistent x1 of eigenvalue barcode scanning information of product, the consistent x2 in appointed place, popularity x3, fraud interests x4 and risk Cost x5 inputs as BP neural network, and target desired output d exports as BP neural network, and BP neural network voluntarily starts to learn Practise training；BP neural network is trained so that BP neural network is built according to multigroup actual product true and false situation data Vertical more accurate.Wherein said output products fraud probability is set as data " 0 " or data " 1 "；Data " 0 " represents fraud probability For 0%, it is 100% that data " 1 " represents fraud probability, is easy to consumer's viewing Query Result.
The present invention sets up BP neural network by the 20 stack features Value Datas providing, and the data in this form is BP nerve net The training data of network, this data combines false proof database for barcode scanning data, and value barcode scanning being obtained through the calculating of S type function fills Change product attribute eigenvalue into.The consistent x1 of barcode scanning information, the consistent x2 in barcode scanning appointed place, popularity is comprised in this form gauge outfit X3, fraud interests x4, value of risk x5 and desired output d, the every data line of form represents the reality whether faked of every set product Border situation data；Application BP neural network algorithm, is simulated training to the product fraud situation of generation, sets up for calculating The BP neural network model of fraud probability, form is as follows：
Wherein, in analog data, using barcode scanning information unanimously as underlying attribute feature, such as fruit product barcode scanning information with anti- The information of pseudo- database purchase can not be completely the same, then eigenvalue is larger, and output fraud probability is also than larger.
Algorithm application module, for carrying out false proof prediction using BP neural network algorithm to product.Establish for calculating After the BP neural network of authenticity of products probability, on the basis of BP neural network, consumer only need to be by barcode scanning device to product two Dimension code is scanned just directly obtaining the true and false probability of product, and it is 0% that data " 0 " represents fraud probability, and data " 1 " represents and makes False probability is 100%, and the method is simply direct, decides whether to buy understand the true and false situation of product according to scanning result after again.
Embodiment 2
As shown in figure 3, the two-dimension code anti-counterfeit Forecasting Methodology based on BP neural network of the present invention, processing procedure includes four Step, implements step as follows：
Step 1：Set up existing false proof database；For accessing product data in false proof database；Enterprise's converted productss The information such as quality information, distributor information, Retailer information, consumer's barcode scanning checking information and packaging Quick Response Code are deposited by data Delivery block is stored in false proof database, and coordinates barcode scanning information to convert the data into the required product genus of BP neural network training Property eigenvalue, beneficial to carrying out to data unifying storing, convenient access data.
Step 2：Scanning obtains product information；By scanning means, product Quick Response Code is scanned, passes through after barcode scanning to sweep Retouch module and analogue signal is converted into digital signal, thus obtaining the information of product.
Step 3：In conjunction with false proof database and barcode scanning information, obtain product attribute eigenvalue；Information after barcode scanning is converted After becoming digital signal data, compare with the product data in false proof database, respectively product attribute eigenvalue is passed through to sweep The code consistent x1 of information, the consistent x2 in appointed place, popularity x3, fraud interests x4 and five aspects of value of risk x5 embody, And assignment is carried out to five product attribute eigenvalues with numeral between [0,1] respectively according to practical situation.Wherein, barcode scanning product Whether identical information information corresponding with false proof database composition barcode scanning information is consistent；Barcode scanning place and formulation selling spot Variation constitute appointed place consistent；Described popularity is led to have differences structure to the prescription of product by living standard difference Become；Different according to the punishment degree that zones of different is faked to product, fraud degree of risk is different, constitutes fraud interests；Because of difference Brand product production cost is different, the differently composed value of risk of product profit height, by this five product attribute eigenvalues Lai True and false prediction is carried out to product so that predict the outcome more accurate；Wherein said product attribute eigenvalue is tied by barcode scanning data Close false proof database input data is normalized, obtain product in [0,1] interval for the numerical value after calculating and belong to Property eigenvalue.
Step 4：BP neural network learning training；Wherein, described BP neural network includes：Input layer, training data is defeated Enter in described input layer；Output layer, generates output from described output layer；And hidden layer, described hidden layer and input layer and Output layer is connected with each other.For BP neural network, the BP nerve net after being trained are trained according to multigroup product attribute eigenvalue Network；BP neural network utilizes the practical situation data of the product whether true and false, is set up by learning training module and can be used for calculating and make The BP neural network of false probability, the self study of BP neural network and adaptive ability are so that predict the outcome relatively accurate.
The specific algorithm of described BP neural network learning training process is as follows：
(1) assign one group of random initial weight to network, its value is between 0 to 1；
(2) by input data normalized, and determine desired output signal d1 according to actual；
(3) real output value successively according to formula calculating, formula is as follows：
In formula
XiThe output valve of i-th node of-input layer；
YjThe output valve of j-th node of-hidden layer；
WijI-th node of-input layer is to the weight coefficient of j-th node of hidden layer；
WjkJ-th node of-hidden layer is to the weight coefficient of k-th node of output layer；
θjThe internal threshold of j-th node of-hidden layer；
θ1The internal threshold of the 1st node of-output layer；
Z1The real output value of the 1st node of-output layer；
Input layer has N number of input node, and hidden layer has M middle layer node, and output layer has 1 output node.
(4) from the beginning of output layer, add momentum α, wherein 0 ＜ α ＜ 1, reversely adjust weights, its adjustment formula is as follows：
Wjk+ηδkYj→Wjk
Wij+ηδjXi→Wij
δj=Yj(1-Yj)·δ1·Wjk
δ in formula1=(d1-Z1)·Z1·(1-Z1).
(5) total error E, i.e. d are calculated1And Z1Between error amount, if E≤ε, wherein ε terminates for study set in advance Absolute error, study stop, otherwise going in step (3) and recalculate.
If wherein step-length η is less, pace of learning is slower, and η causes network to occur swinging compared with conference, then in step (4) Add a momentum α, wherein 0 ＜ α ＜ 1, that is,
Wjk+ηδkYj+α·ΔWjk→Wjk
Wij+ηδjXi+α·ΔWij→Wij
Δ W in formulajk- double WjkDifference；
ΔWij- double WijDifference.
Step 5：Using BP neural network algorithm, false proof prediction is carried out to barcode scanning product；Establish for calculating authenticity of products After the BP neural network of probability, consumer only need to be scanned to product Quick Response Code by barcode scanning device, carries to BP neural network For input signal x1, produce output signal Z in output layer after calculating, just can directly obtain the true and false probability of product；Should Method is simply direct, decides whether to buy understand the true and false situation of product according to scanning result after again.
The Antiforge system of present invention design, false proof database and BP neural network algorithm is combined, in theory and practice All there is very big value, the probability of authenticity of products can not only be predicted, authenticity of products is judged avoid thinking in absolute terms, carry out Rational judgment, avoids the erroneous judgement of authenticity of products to a certain extent；Quick Response Code is placed in outside product simultaneously, can allow consumer First barcode scanning judges the true and false, then decides whether to buy, and so can solve the problem that consumer must first buy the defect of product ability barcode scanning, In addition, checking authenticity of products before shopping, meet consumer's shopping psychology demand, promote purchase intention, have in actual applications very Big value.
It will be apparent to those skilled in the art that can technical scheme as described above and design, make other various Corresponding change and deformation, and all these change and deformation all should belong to the protection domain of the claims in the present invention Within.

Claims (9)

1. two-dimension code anti-counterfeit prediction meanss based on BP neural network are it is characterised in that include being sequentially connected by signal flow direction Data access module, barcode scanning module, data analysis module, learning training module and algorithm application module；Described BP neural network It is made up of input layer, hidden layer and output layer；
Data access module, sets up false proof database for accessing product data；
Barcode scanning module, for obtaining barcode scanning data；
Data analysis module, for calling false proof database that the data that barcode scanning obtains is carried out with Treatment Analysis, obtains the genus of product Property eigenvalue；
Learning training module, the characteristic value data being obtained by actual product true and false situation is faked general as network inputs, product Rate, as output, constantly changes the connection weight of network under the stimulation of input, so that the output of network is constantly close to expectation Output, thus constructing BP neural network algorithm；
Algorithm application module, carries out false proof prediction using BP neural network algorithm to barcode scanning product.
2. the two-dimension code anti-counterfeit prediction meanss based on BP neural network according to claim 1 are it is characterised in that described number It is used for the converted productss quality information of enterprise, distributor information, Retailer information, consumer's barcode scanning checking letter according to access module Breath and packaging 2 D code information are all stored and are set up false proof database.
3. the two-dimension code anti-counterfeit prediction meanss based on BP neural network according to claim 1 are it is characterised in that described product Product attribute character value includes the consistent x1 of barcode scanning information, the consistent x2 in barcode scanning appointed place, popularity x3, fraud interests x4, risk generation Valency x5；The information of barcode scanning product information corresponding with the false proof database whether identical composition consistent x1 of barcode scanning information；Barcode scanning Place constitutes the consistent x2 in barcode scanning appointed place with the variation in designated sale place；Described popularity x3 is led to by living standard difference Composition is had differences to the prescription of product；Different, the fraud risk journey according to the punishment degree that zones of different is faked to product Degree is different, constitutes fraud interests x4；Because different brands production cost is different, the differently composed value of risk of product profit height x5.
4. the two-dimension code anti-counterfeit prediction meanss based on BP neural network according to claim 1 are it is characterised in that described number Using S type function, the data that barcode scanning obtains is normalized with reference to false proof database according to analysis module, barcode scanning is obtained Value be converted into [0,1] interval in, provide initial data for the study for BP neural network；Wherein S type function formula is：
Input
Net=x1, w1+w2, w1w2+…+xnwn
Output
y = f ( n e t ) = 1 1 + e - n e t
Wherein xnFor n-th input signal；WnFor weight coefficient.
5. the two-dimension code anti-counterfeit prediction meanss based on BP neural network according to claim 1 are it is characterised in that described Practising training module is according to input signal XiOutput node is used as by intermediate node, through non-linear transformations, produces output letter Number Zk, each sample of network training includes input vector X and desired throughput d, network output valve Z and desired output d it Between deviation, by adjusting connection weight W of input node and hidden nodeijConnection and hidden node and output node between Intensity TjkAnd threshold value, so that error is declined along gradient direction, through repetition learning training, determine the net corresponding with minimum error Network parameter, i.e. weights and threshold value, training terminates.
6. the two-dimension code anti-counterfeit Forecasting Methodology based on BP neural network is it is characterised in that include：
Step 1：Set up existing false proof database；
Step 2：Barcode scanning obtains product information；
Step 3：In conjunction with false proof database and barcode scanning information, obtain product attribute eigenvalue；
Step 4：BP neural network learning training；
Step 5：Using BP neural network algorithm, false proof prediction is carried out to barcode scanning product.
7. the two-dimension code anti-counterfeit Forecasting Methodology based on BP neural network according to claim 6 is it is characterised in that described step Product attribute eigenvalue described in rapid 3 is normalized to input data with reference to false proof database by barcode scanning data, warp Cross the product attribute eigenvalue obtaining numerical value after calculating in [0,1] interval.
8. the two-dimension code anti-counterfeit Forecasting Methodology based on BP neural network according to claim 6 is it is characterised in that described step The specific algorithm of the BP neural network learning training process in rapid 4 is as follows：
(1) assign one group of random initial weight to network, its value is between 0 to 1；
(2) by input data normalized, and determine desired output signal d1 according to actual；
(3) real output value successively according to formula calculating, formula is as follows：
Y j = f ( Σ i = 0 N - 1 W i j X i - θ j ) , j = 0 , 1 , ... , M - 1
Z 1 = f ( Σ j = 0 M - 1 W j k Y j - θ 1 )
In formula
XiThe output valve of i-th node of-input layer；
YjThe output valve of j-th node of-hidden layer；
WijI-th node of-input layer is to the weight coefficient of j-th node of hidden layer；
WjkJ-th node of-hidden layer is to the weight coefficient of k-th node of output layer；
θjThe internal threshold of j-th node of-hidden layer；
θ1The internal threshold of the 1st node of-output layer；
Z1The real output value of the 1st node of-output layer；
Input layer has N number of input node, and hidden layer has M middle layer node, and output layer has 1 output node.
(4) from the beginning of output layer, addition momentum α, wherein 0<α<1, reversely adjust weights, its adjustment formula is as follows：
Wjk+ηδkYj→Wjk
Wij+ηδjXi→Wij
δj=Yj(1-Yj)·δ1·Wjk
δ in formula1=(d1-Z1)·Z1·(1-Z1).
(5) total error E, i.e. d are calculated1And Z1Between error amount, if E≤ε, wherein ε be set in advance study terminate exhausted To error, study stops, otherwise going in step (3) and recalculate.
If 9. the two-dimension code anti-counterfeit Forecasting Methodology based on BP neural network according to claim 6 or 8 it is characterised in that The less then pace of learning of step-length η is slower, and η causes network to occur swinging compared with conference, then add a momentum α in step (4), Wherein 0 ＜ α ＜ 1, that is,
Wjk+ηδkYj+α·ΔWjk→Wjk
Wij+ηδjXi+α·ΔWij→Wij
Δ W in formulajk- double WjkDifference；
ΔWij- double WijDifference.
CN201610843520.8A 2016-09-21 2016-09-21 Two-dimension code anti-counterfeit prediction meanss and method based on BP neural network Active CN106469336B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610843520.8A CN106469336B (en) 2016-09-21 2016-09-21 Two-dimension code anti-counterfeit prediction meanss and method based on BP neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610843520.8A CN106469336B (en) 2016-09-21 2016-09-21 Two-dimension code anti-counterfeit prediction meanss and method based on BP neural network

Publications (2)

Publication Number Publication Date
CN106469336A true CN106469336A (en) 2017-03-01
CN106469336B CN106469336B (en) 2018-04-17

Family

ID=58230524

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610843520.8A Active CN106469336B (en) 2016-09-21 2016-09-21 Two-dimension code anti-counterfeit prediction meanss and method based on BP neural network

Country Status (1)

Country Link
CN (1) CN106469336B (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108520155A (en) * 2018-04-11 2018-09-11 大连理工大学 Vehicle behavior analogy method based on neural network
CN109214193A (en) * 2017-07-05 2019-01-15 阿里巴巴集团控股有限公司 Data encryption, machine learning model training method, device and electronic equipment
CN109766944A (en) * 2019-01-10 2019-05-17 四川中新华搜信息技术有限公司 A kind of exterior of commodity method for anti-counterfeit of the image recognition based on CNN
CN109887202A (en) * 2019-02-22 2019-06-14 北京航天泰坦科技股份有限公司 The tax control tray of built-in enhancing module
WO2019246473A1 (en) * 2018-06-21 2019-12-26 The Procter & Gamble Company Method of determining authenticity of a consumer good
CN111680772A (en) * 2020-06-02 2020-09-18 云南省烟草质量监督检测站 Method, device and equipment for detecting authenticity of cigarette strip and box packaging paper
CN114943546A (en) * 2022-05-25 2022-08-26 南京数睿数据科技有限公司 Article anti-counterfeiting detection method and device, electronic equipment and computer readable medium
CN117411615A (en) * 2023-12-13 2024-01-16 广州市信亦达电子科技有限公司 Two-dimensional code anti-counterfeiting encryption method and system based on random number

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7181493B2 (en) * 2003-12-23 2007-02-20 Unisys Corporation Platform independent model-based framework for exchanging information in the justice system
CN101383028A (en) * 2008-10-06 2009-03-11 四川正道天和信息科技有限公司 National commodity electronic monitoring method based on EPC article networking and system thereof

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7181493B2 (en) * 2003-12-23 2007-02-20 Unisys Corporation Platform independent model-based framework for exchanging information in the justice system
CN101383028A (en) * 2008-10-06 2009-03-11 四川正道天和信息科技有限公司 National commodity electronic monitoring method based on EPC article networking and system thereof

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
肖美芳: "《基于二维码技术的白酒防伪管理研究》", 《中国优秀硕士学位论文全文数据库(电子期刊)》 *
陈贵川: "《基于数据挖掘的二维码防伪防窜货系统分析与设计》", 《中国优秀硕士学位论文全文数据库（电子期刊）》 *

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109214193A (en) * 2017-07-05 2019-01-15 阿里巴巴集团控股有限公司 Data encryption, machine learning model training method, device and electronic equipment
CN109214193B (en) * 2017-07-05 2022-03-22 创新先进技术有限公司 Data encryption and machine learning model training method and device and electronic equipment
CN108520155A (en) * 2018-04-11 2018-09-11 大连理工大学 Vehicle behavior analogy method based on neural network
WO2019246473A1 (en) * 2018-06-21 2019-12-26 The Procter & Gamble Company Method of determining authenticity of a consumer good
US11074592B2 (en) 2018-06-21 2021-07-27 The Procter & Gamble Company Method of determining authenticity of a consumer good
CN109766944A (en) * 2019-01-10 2019-05-17 四川中新华搜信息技术有限公司 A kind of exterior of commodity method for anti-counterfeit of the image recognition based on CNN
CN109887202A (en) * 2019-02-22 2019-06-14 北京航天泰坦科技股份有限公司 The tax control tray of built-in enhancing module
CN111680772A (en) * 2020-06-02 2020-09-18 云南省烟草质量监督检测站 Method, device and equipment for detecting authenticity of cigarette strip and box packaging paper
CN114943546A (en) * 2022-05-25 2022-08-26 南京数睿数据科技有限公司 Article anti-counterfeiting detection method and device, electronic equipment and computer readable medium
CN114943546B (en) * 2022-05-25 2024-04-09 南京数睿数据科技有限公司 Article anti-counterfeiting detection method, device, electronic equipment and computer readable medium
CN117411615A (en) * 2023-12-13 2024-01-16 广州市信亦达电子科技有限公司 Two-dimensional code anti-counterfeiting encryption method and system based on random number
CN117411615B (en) * 2023-12-13 2024-04-02 广州市信亦达电子科技有限公司 Two-dimensional code anti-counterfeiting encryption method and system based on random number

Also Published As

Publication number Publication date
CN106469336B (en) 2018-04-17

Similar Documents

Publication Publication Date Title
CN106469336B (en) Two-dimension code anti-counterfeit prediction meanss and method based on BP neural network
CN107230113A (en) A kind of house property appraisal procedure of multi-model fusion
Morduch Consumption smoothing across space: Testing theories of risk-sharing in the ICRISAT study region of South India
CN105447235B (en) A kind of soil future utilizes scene Dynamic Simulation Method
CN103577888B (en) Method for optimization of product design schemes
Burkett The value problem in ecological economics: lessons from the physiocrats and Marx
CN102982229B (en) The data preprocessing method of a kind of multi items price forecasting of commodity based on neural network
CN109117982A (en) A kind of room rate prediction technique combined based on deep learning with feature
CN110020868A (en) Anti- fraud module Decision fusion method based on online trading feature
Jiannan et al. Impacts of the choice habits of port users on the effects and efficiencies of port investment
CN108876480A (en) Large agricultural product price trend forecasting method based on multi-source data
Fu et al. Applying DEA–BPN to enhance the explanatory power of performance measurement
CN104484580B (en) Antibacterial peptide Activity Prediction method based on Multi-label learning
Cumming Mortgages, cash-flow shocks and local employment
Denysiuk et al. Diversification in the enterprises' activities for sustainable development in the agricultural sector
CN108846341A (en) A kind of remote sensing images lake ice classifying identification method neural network based
CN108052755A (en) Vector space based on completely random forest calculates intensity prediction method and system
Chen Construction and Application of Forecast Model for Eco-Tourism Market Demand Based on BP Neural Network
Adelman et al. Institutions and development in the nineteenth century: A latent variable regression model
CN108460614A (en) Property price automated decision system based on integrated learning approach and method
CN115983142B (en) Regional population evolution model construction method based on depth generation countermeasure network
Deng et al. Analysis of the environmental Kuznets curve for forest fragmentation: The case of Beijing-Tianjin-Hebei region in China
Wang et al. Research on Group Risk Security Decision Based on BP Neural Network Algorithm
Avianto et al. Development of spatial-system dynamics model for food security policy in Indonesia: A generic sub-model simulation
KR20230081028A (en) Method for generating a product demand forecasting model and for predicting product demand using the same

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant